Skip to main content

Table 6 Performance of predictive, prognostic and non-personalized models based on out-of-sample and in-sample predictions

From: Framework for personalized prediction of treatment response in relapsing remitting multiple sclerosis

Measure

Out-of-sample mean (SE)a

In-sample mean (SE)a

Sample sizeb

Response

Model

C-Index

0.5819 (0.0008)

0.6546 (0.0005)

307,784

CDP

predictive

C-Index

0.5625 (0.0007)

0.6220 (0.0004)

307,784

CDP

prognostic

C-Index

0.5467 (0.0006)

0.5649 (0.0005)

307,784

CDP

non-personalized

C-Index

0.6458 (0.0004)

0.6781 (0.0003)

505,724

relapse

predictive

C-Index

0.6482 (0.0003)

0.6700 (0.0002)

505,724

relapse

prognostic

C-Index

0.5531 (0.0003)

0.5609 (0.0003)

505,724

relapse

non-personalized

MSE

0.12497 (0.00005)

0.11928 (0.00004)

3119

CDP

predictive

MSE

0.12486 (0.00004)

0.12132 (0.00003)

3119

CDP

prognostic

MSE

0.12449 (0.00002)

0.12388 (0.00001)

3119

CDP

non-personalized

MSE

0.7554 (0.0008)

0.7097 (0.0006)

3119

relapse

predictive

MSE

0.7312 (0.0005)

0.7049 (0.0003)

3119

relapse

prognostic

MSE

0.7557 (0.0002)

0.7517 (0.0001)

3119

relapse

non-personalized

NLL

1252.6 (0.6)

1190.9 (0.4)

3119

CDP

predictive

NLL

1254.0 (0.5)

1215.5 (0.2)

3119

CDP

prognostic

NLL

1246.8 (0.2)

1240.3 (0.1)

3119

CDP

non-personalized

NLL

2580.8 (0.6)

2519.9 (0.4)

3119

relapse

predictive

NLL

2574.6 (0.5)

2534.9 (0.3)

3119

relapse

prognostic

NLL

2650.1 (0.2)

2641.9 (0.2)

3119

relapse

non-personalized

  1. SE standard error of the mean, CDP confirmed disability progression, C-Index Harrell’s concordance statistic, MSE mean squared error, NLL negative log-likelihood.
  2. a Estimated by repeating 10-fold cross-validation 40 times
  3. b Refers either to the number of observations (MSE, NLL) or the number of matched pairs (C-Index)